Last week, Anthropic's announcement that it had added a text watermark to Claude models drew considerable attention. This watermark covers every Claude AI surface - even text generated through direct API calls carries the same detectable signature. Notably, this is not a watermark built on inserting special Unicode characters. Instead, it operates through a fundamentally different mechanism: word-level substitution. How the Watermark Actually Works…

To comply with the transparency regulations stipulated by the European Union Artificial Intelligence Act (EU AI Act), Anthropic recently disclosed the underlying mechanics of its text watermarking system. You can explore the comprehensive technical details regarding the Claude text watermark directly on Anthropic's official news portal. Unlike the commonly imagined hidden code or visual markers, Anthropic employs an exceedingly covert statistical algorithm.

#Claude’s #textwatermark, implemented to comply with the #EU #AIAct, uses a method that doesn’t impact the quality or content of outputs. The #watermark, undetectable to readers, is based on low-stakes word choices and can be detected using a key, indicating the likelihood of Claude’s involvement in generating the text. https://www.anthropic.com/news/claude-text-watermark?eicker.news #tech #news #ainews

来来，试试我新出炉的word混淆API，可查找，但无法好好的复制整段的文字到剪贴板。 在逗号和句号后，会加上自定义的字符及随机的字符，复制整段，保证欲仙俗死。 http://api.r34.cc/jb/textwatermark 下面文档是我转化后的效果，有兴趣可以看看区别。 pdf版本 pub.r34.cc ### wordwater.pdf 610.36 KB 原始稿及pdf等 https://pub.r34.cc/pic/wordwater.zip